PASS
The requirements for the exact governed decision are satisfied under the evaluated evidence, authority and policy.
PASS does not itself execute an external action.
VERIFIABLE TRUST INFRASTRUCTURE
Verifiable trust infrastructure
for consequential decisions.
VellumOS is verifiable trust infrastructure for consequential decisions, built around explicit evidence, authority, policy, human review, scoped permission, controlled execution, and reconstructable result evidence.
The objective is not simply to automate a workflow.
It is to make the basis of a consequential decision reviewable and reconstructable.
THE PROBLEM
Consequential decisions break down
when their decision basis is implicit.
In regulated and high-consequence environments, the difficult part is often not moving work from one step to another.
The difficult part is determining whether a specific action may advance, based on the right evidence, under the right policy, with the right authority, while preserving a record that can later be reconstructed and reviewed.
Faster workflows do not automatically create more defensible decisions.
WHO IT'S FOR
For organizations accountable for consequential decisions.
VellumOS is being developed for organizations that must repeatedly decide whether a regulated or consequential action may proceed.
They evaluate evidence, identifiable authority, and applicable policy while keeping that basis reconstructable, with human review where ambiguity or exceptions require it.
Delays, unresolved exceptions, and missing evidence can compromise timely decisions and make their basis harder to reconstruct. VellumOS addresses that recurring bottleneck as governed decision infrastructure—not a document manager, form product, or generic workflow engine.
ONE COHERENT TRUST SYSTEM
Trust is not a collection of features.
It is a system.
VellumOS is designed as one coherent trust infrastructure. Evidence, authority, policy, decision, human review, permission, execution, and recorded result are distinct responsibilities, connected through explicit, governed, versioned contracts.
Each component has a defined responsibility. Their relevant interactions are governed and traceable, while authority, access, and failure remain within defined boundaries.
THE TRUST PATH
From evidence to a reconstructable decision.
Domain-specific decision contracts define the evidence, authority, and policy required for each decision.
Each stage has a distinct responsibility.
Evidence does not become authority.
A favorable decision does not itself execute an external action.
Execution does not erase the need to preserve the basis and result of the decision.
- 01
Evidence
Evidence is admitted with provenance, purpose and context.
- 02
Authority
The system establishes who or what is authorized to participate in the governed decision.
- 03
Policy
Applicable rules, constraints and decision conditions are evaluated deterministically where required.
- 04
Decision
The exact governed question is assessed against the available evidence, authority and policy.
- 05
Human Review
where requiredAmbiguity, exceptions or conditions requiring judgment remain reviewable by an authorized human.
- 06
Scoped Permission
A favorable assessment may support a bounded permission for an exact action.
Permission is not the same as execution.
- 07
Controlled Execution
An authorized action may proceed only within its defined scope and execution boundary.
- 08
Result Evidence
The system records evidence about the execution attempt and its reported result without automatically treating a receipt as external-world truth.
- 09
Reconstruction
A later reviewer can reconstruct the decision path from its evidence, authority, policy, review, permission and recorded result.
Independent Verification / Portable Review
DECISION OUTCOMES
A governed decision resolves to one of three states.
REVIEW
Human review is required before the governed decision can advance under the current conditions.
The review remains bounded to the applicable evidence, authority, policy and decision context.
BLOCK
The governed action is not authorized to advance under the current conditions.
BLOCK does not claim that the external-world action is physically impossible through another path.
VERIFIABLE EVIDENCE
Evidence retains the context
in which it may be used.
VellumOS is designed to preserve evidence provenance, integrity, purpose and decision context so a reviewer can understand what was admitted, how it was used, and what role it played in a governed decision.
A hash, receipt or provenance record can support bounded integrity of a recorded artifact.
It does not, by itself, establish external-world truth.
PORTABLE TRACEABILITY
The decision path should remain
reviewable beyond one interface.
The evidence supporting a consequential decision should not depend on operator memory or on a single interface remaining available.
VellumOS is designed as complementary trust infrastructure that can sit alongside systems organizations already operate. It is not those systems of record, and it does not take their decision authority.
Portable evidence may support later review in another context only when purpose, authority, policy, and permission remain explicit. That is not live integration with incumbent platforms, and it is not unrestricted sharing across organizations.
PORTABLE EVIDENCE
Operator-independent verification
Decisions remain reviewable without depending on VellumOS as the operator.
Portable decision evidence allows an authorized reviewer to reconstruct the governed path from preserved evidence, authority, policy, review, permission and recorded result.
HUMAN REVIEW
Human judgment remains visible
where judgment is required.
Human review is part of the architecture when deterministic automation should not resolve ambiguity, exceptions or consequential judgment.
Review remains scoped to the applicable decision, evidence, authority and policy rather than acting as an unrestricted override.
AI IN A GOVERNED CONTEXT
Probabilistic assistance.
Governed authority.
AI is designed to assist interpretation, extraction, classification, and other bounded tasks. It does not become decision authority.
AI output does not automatically become evidence, authority, or permission to act. Consequential boundaries remain governed by purpose, evidence, policy, authority, and human review where required.
Probabilistic intelligence can assist a governed process.
It should not silently inherit decision authority. This page does not claim live external-model operations.
CONTROLLED EXECUTION
Permission is explicit
before execution.
Decision authority and execution authority are separate concerns.
A favorable decision assessment may support a narrowly scoped permission for an exact action, but PASS is not itself execution.
Controlled execution is the architectural boundary between an authorized decision and an external effect.
A decision assessment is not scoped permission. Permission is not an execution attempt. A recorded result is not external-world finality. PASS does not execute. This page does not claim live external effects.
ARCHITECTURAL PRINCIPLES
Trust should not depend
on an operator’s assertion.
Compliance-First is a foundational architectural principle: controls, evidence, and claims discipline come before features. It is not a SOC 2, HIPAA, or GDPR certification claim.
Verifiable integrity
Recorded artifacts and decision evidence should support independent integrity checks.
Explicit authority
Authority must be identifiable and bounded to the applicable decision context.
Versioned policy
The applicable policy and decision conditions must remain identifiable and reconstructable.
Human-reviewable decision paths
Human judgment must remain visible where automation should not resolve ambiguity.
Auditable evidence
A reviewer should be able to understand what evidence participated in a decision and why.
Scoped execution
Permission to act must remain bounded to the exact authorized action and context.
Portable traceability
Decision evidence should remain reconstructable without depending on informal operator memory.
Operator-independent trust
Confidence should derive from verifiable evidence and system controls rather than assertion by the system operator alone.
RESPONSIBLE BOUNDARIES
Precision includes stating
what has not been claimed.
VellumOS is under controlled development.
This page does not describe a generally available product or a catalog of industry deployments.
What this page describes is architecture under controlled development. Internal and synthetic proofs are not customer deployments, live incumbent integrations, Production operations, or demonstrated operating effectiveness.
This website is not a deployed public verifier. A hash or receipt can support bounded integrity of a recorded artifact; it does not, by itself, establish external-world truth.
This institutional website does not represent general availability, a Production operating claim, government or agency approval, direct government filing capability, or certification under SOC 2, HIPAA or GDPR.
It does not claim guaranteed court admissibility, implemented blockchain, implemented zero-knowledge proofs, a deployed public verifier, or universal external-world finality.
Public institutional presence is not Public GTM and is not a customer onboarding program.
COMPANY
Infrastructure for decisions
that must withstand scrutiny.
VellumOS, Inc. develops verifiable trust infrastructure for consequential and regulated decisions.
The architecture is being developed around explicit evidence, authority, policy, human review, controlled execution and reconstructable decision evidence.